When an AI Agent Shops for Software, Your Own Agent Negotiates the Deal.

haggl.ai is the negotiation layer for the agentic web. Buyers increasingly send an AI agent to pick their software — it reads pricing pages, compares alternatives, and recommends one. You install a single script tag and describe your business: who your most valuable customers are and your published plans. From then on, an AI negotiator represents you in every session an agent opens. The buyer’s agent gets a personalized deal backed by verified evidence, and closes on a dedicated Stripe Checkout link at the negotiated price. You get a customer you chose to win, at a discount you authorized — with no coupon code in circulation.

A frontier-model AI agent, briefed with your business description, your published plans, and your customer segments. It negotiates each session with judgment — weighing the evidence an agent presents, countering, asking for more — rather than running a rules engine. Code enforces exactly one hard limit: the maximum discount you set per segment. The negotiator can be creative below that ceiling; it can never cross it.

A buyer agent discovers your negotiation endpoint and fetches it. The response documents the whole protocol: your customer segments, the request schema, and how negotiation works — agents learn everything from the API itself. The agent makes a free opening offer with self-reported data (team size, use case, current tool, renewal date), then strengthens its case with evidence: real emails from its user’s inbox — invoices from the incumbent, seat reports, renewal notices — submitted with the user’s consent. haggl.ai verifies each email’s origin cryptographically before your negotiator prices the next offer. When both sides agree, the agent receives a dedicated Stripe Checkout link for the plan that segment maps to, at the negotiated price. Every step is structured and fully logged.

A promo code is a shared secret: once it exists, it spreads — coupon sites, Slack communities, customers who would have paid full price. A negotiated deal on haggl.ai becomes a Stripe Checkout Session created on your own Stripe account: one link, one buyer, the exact negotiated percentage, valid for 24 hours and consumed on payment. There is no code to paste, forward, or leak, and the discount is scoped to the plan you mapped to that segment. Nothing changes for your public pricing page.

Yes. Connect your Stripe account with Stripe Connect (OAuth — haggl.ai acts on your account, never holds your keys), then pick which Stripe price each customer segment maps to: a monthly or annual subscription, or a one-time price. On accept, haggl.ai creates the Checkout Session in your account with the negotiated discount applied. Not on Stripe? Give each segment a checkout URL of your own instead and the negotiator sends buyers there.

The standard hosted flow checks email signatures (DKIM) to establish signed origin, not every claim about the customer. Original message fields are removed before pricing-model input and storage; submitted excerpts, profile fields and verification results may remain. Custom integrations need their own data review.

Agents discover you three ways: a machine-readable meta tag the embed injects into your page, WebMCP tools registered in the browser, and haggl.ai’s MCP server. No buyer-side install is needed — the endpoint documents itself in plain language written for an LLM to parse. In our own August 2026 testing on production consumer apps, ChatGPT, Claude, and Grok all detected the negotiation capability on a demo vendor’s site and completed real negotiations; Gemini did not yet surface it. As agentic procurement grows, haggl-enabled pricing pages are the ones agents can actually do business with.

Because negotiating costs agents nothing and consistently produces better outcomes for their users. Unlike humans — who find negotiation awkward, slow, or uncertain, and who never fill in “Contact sales” — agents don’t hesitate. The protocol makes the first move free: an opening offer on self-reported data shares nothing personal and always returns a real quote. An agent that can get its user 20% off your Business plan will recommend you over a competitor whose pricing page is a dead end.

One script tag on your marketing site or app, then describe your business in the dashboard: who your valuable customer segments are, the maximum discount each is worth, and your published plans. Connect Stripe and map each segment to a plan, or paste your own checkout URLs. That’s the entire setup — no SDK, no backend changes, no billing integration to write. Your AI negotiator is live the moment you finish, typically well under fifteen minutes.

Every plan starts free: your first 50 negotiations each month cost nothing, no credit card required. Paid plans scale with volume — Starter at $49/month (300 negotiations), Growth at $149/month (1,000), and Pro at $399/month (5,000). Enterprise pricing for high-volume or custom deployments is available on request.

The customer asks their AI agent to find the right tool at the best price. The agent discovers your endpoint, negotiates autonomously, and — before sharing anything personal — asks its user for consent, often with an explicit yes/no prompt. Then it returns with a confirmed offer and a checkout link that already carries the negotiated price. No form, no sales call, no code to type. From the customer’s perspective, their agent went shopping and came back with a better deal than the pricing page.

Yes — consent is a designed moment in the protocol, not fine print. The opening offer uses only self-reported data, so nothing personal is needed to start. When evidence would improve the deal, the agent asks its user before submitting anything, and you see nothing until the user says yes. Even then, haggl.ai verifies the evidence and stores only cryptographic hashes — never the underlying emails.

Verified facts about who the customer is, plus the full negotiation record. Your segments should describe lifetime value — seat expansion, multi-year retention, low cost-to-serve, whether they refer others — because that’s what the discount is buying. The segments are your discount-authorization framework: agents proving high-LTV fit unlock deeper discounts; everyone else gets a thinner one. Purchase intent isn’t the question — the agent is already at the table.

Every negotiation is logged in your haggl dashboard with the verified submission, the offer history, and the outcome, and you get a deal email the moment an offer is accepted. You can also configure a webhook to push structured lead data to your CRM in real time. Payments land in your own Stripe account like any other checkout.

Agents see only what the protocol publishes: your public segment descriptions, the maximum discount each authorizes, and the plans you’ve chosen to publish. Your private strategy notes — why a segment matters, how you think about its value — never leave the server, and your negotiator quotes only figures written in your brief. Checkout links are created per deal, so no standing discount exists to be discovered.

ACP (OpenAI + Stripe) and UCP (Google, Shopify, Walmart) handle product discovery and checkout — they answer “what do you sell?” and “how do I pay?” haggl.ai handles the step neither protocol covers: negotiation. It answers “what deal can I get based on who I am?” The three are complementary — vendors can run ACP or UCP for discovery and checkout, with haggl setting the personalized price in between.

zkTLS will let an agent cryptographically prove that data came from a specific source — a competitor’s admin panel, a billing dashboard, a usage report — without exposing the raw content. It extends haggl.ai’s verified-evidence model beyond the inbox: today origin is proven with cryptographic email-signature checks (DKIM), live in production; zkTLS adds first-party portals next. Support is coming in an upcoming release.